LGALS1 drives chemoresistance in esophageal squamous cell carcinoma by modulating epithelial-mesenchymal transition and tumor immunity

Abstract

BACKGROUND

Esophageal cancer represents one of the most prevalent malignant tumors globally. Esophageal cancer lack of tumor markers in clinical diagnosis, unable to effectively monitor development, drug resistance and prognosis of tumor, lead to clinical treatment effect is poorer. Neoadjuvant chemoradiotherapy (neoCRT) can substantially enhance the prognosis for individuals diagnosed with locally advanced esophageal squamous cell carcinoma (ESCC). Nevertheless, treatment resistance still occurs, affecting the long-term survival of patients.

AIM

To investigate the genes and regulatory mechanisms associated with platinum-based resistance in ESCC.

METHODS

In this investigation, single-cell RNA sequencing results of ESCC were analyzed to obtain the dynamic remodeling of tumor microenvironment in ESCC patients following neoCRT. Six ESCC samples were analyzed for scRNA-seq analysis. Four patients achieved partial response after neoCRT and then underwent surgery, and 2 patients received surgery alone. Multiple immunofluorescence and western blot were used to verify the expression of characteristic genes and the distribution of tumor cells. Gene knockdown, cholecystokinin-8, flow cytometry, colony formation, and subcutaneous tumorigenesis were used to validate the role of signature genes in the development of platinum-resistant ESCC.

RESULTS

Analysis showed that individuals with ESCC after neoadjuvant chemoradiation, expressed in the tumor cell toxicity molecular effect of CD8+ T cell infiltration, inflammation macrophages subgroup abundance increase, LAMP3+ dendritic cell activation, increase antitumor immune response. However, some residual malignant epithelial cells still survived, indicating that these cells may have drug resistance and the possibility of relapse. We analyzed this part of the cell gene expression, and screening to LGALS1 gene may be associated with drug resistance. LGALS1 exhibits elevated expression levels in tumor cells and is linked to cisplatin resistance. Functional verification indicated that knockdown of LGALS1 expression could up-regulate the sensitivity of esophageal squamous cell cells and tumors to cisplatin therapy by inhibiting intracellular epithelial-mesenchymal transformation, DNA damage repair and anti-apoptosis mechanisms.

CONCLUSION

These findings confirm LGALS1 is the key of ESCC of platinum resistance protein, targeted LGALS1 may be an effective means to overcome the neoadjuvant chemoradiation resistance.

Key Words: Neoadjuvant chemoradiotherapy; Esophageal squamous cell carcinoma; Single-cell RNA sequencing; LGALS1; Cisplatin resistance

Core Tip: In this investigation, we comprehensively characterized the dynamic changes in the immune microenvironment during treatment by analyzing single-cell RNA sequencing data of esophageal squamous cell carcinoma (ESCC) patients who underwent neoadjuvant chemoradiotherapy. LGALS1 was found to be a gene associated with neoadjuvant chemoradiotherapy resistance in ESCC. Subsequent experimental validation demonstrated LGALS1’s contribution to cisplatin resistance in ESCC through both in vitro and in vivo investigations. These studies illustrate ESCC cisplatin drug treatment failure mechanisms, emphasized the LGALS1 can serve as a promising therapeutic target. It also reveals a potential strategy to overcome the development of cisplatin resistance in ESCC, laying a foundation for the development of more effective individualized treatment regimens, and potentially leading to better outcomes for patients with advanced esophageal cancer.

INTRODUCTION

Esophageal cancer (EC) exhibits high invasiveness and represents a malignant tumor with increased incidence and mortality rates worldwide, ranking seventh and sixth respectively among all cancers globally[1]. EC is primarily split into esophageal adenocarcinoma and esophageal squamous cell carcinoma (ESCC), and there are large regional differences. Among them, esophageal adenocarcinoma mainly occurs in Europe and North America, and ESCC is dominant in Asia[2]. Due to the great concealment of its occurrence and development, most of the patients are diagnosed in the local advanced stage, resulting in an overall unfavorable prognosis. The hypoxic environment within the tumor promotes angiogenesis and metastasis, and effectively down-regulates tumor sensitivity to radiotherapy and chemotherapy. For the treatment of ESCC, surgery, chemotherapy, radiotherapy and immunotherapy still have limitations, leading to recurrence and tumor resistance[3]. The development of more precise and less toxic therapies is urgently needed.

Clinically, neoadjuvant chemoradiotherapy (neoCRT) combined with surgical treatment has demonstrated notable enhancement in survival outcomes for individuals with locally advanced EC[4,5]. Recently, the emergence of immune-based treatments, particularly the integration of checkpoint blockade agents with cytotoxic therapy, has yielded substantial improvements in achieving pathological complete remission (pCR) among EC patients[6]. Although clinical progress has been made, the majority of patients will still experience recurrence, which will still lead to the death of patients with end-stage tumors, indicating the urgent need to address the treatment resistance of ESCC[7]. The development of neoCRT resistance in ESCC is related to the interaction of epithelial-mesenchymal transition (EMT), Notch, Wnt and Nrf2 signaling pathways[8], as well as immune cells in tumor microenvironment (TME)[9]. The mechanism of resistance to neoCRT is complex and unclear, which needs further study.

LGALS1, also known as Galectin-1, belongs to the β-galactoside-binding soluble lectin family. Intratumoral LGALS1 can promote tumor escape from immune surveillance and promote tumor growth[10]. In pancreatic cancer, LGALS1 occurs in high expression and is significantly associated with poor overall survival in pancreatic cancer[11]. In gastrointestinal malignancies, LGALS1 promotes tumor development, metastasis, immunosuppression and angiogenesis[12]. In hepatocellular carcinoma (HCC), the expression of LGALS1 is up-regulated in HCC cells compared with normal cells, and LGALS1 is associated with HCC cell migration and invasion, and is associated with tumor growth, invasion, metastasis, postoperative recurrence and poor prognosis[13]. LGALS1 can protect head and neck squamous cell carcinoma and inhibit the killing effect of immune system on neck squamous cell carcinoma[14].

Single-cell RNA sequencing (scRNA-seq) has become an effective tool to analyze cancer development and drug resistance. By analyzing the heterogeneity and dynamic distribution of cell subsets in the TME before and after treatment, scRNA-seq provides new insights into tumor evolution and treatment[15,16]. ScRNA-seq can accurately detect resistance-related gene changes and signaling pathway dysregulation in various cell subsets, providing us with new molecular architectures of ESCC treatment failure. To screen therapeutic targets in single cells within ESCC tumors by scRNA-seq, perform functional verification and rigorous preclinical trials, and propose novel approaches for developing efficacious treatment modalities. These efforts will ultimately contribute to the clinical improvement of personalized treatment methods for ESCC and enhance the clinical outcomes of ESCC[17].

MATERIALS AND METHODS

Acquisition and analysis of scRNA-seq data

ScRNA-seq data of 6 ESCC patients were obtained from the GSE221561 dataset[18]. Four patients achieved partial response after neoCRT and then underwent surgery, and 2 patients only underwent surgery. The data included tumor tissues and adjacent tissues. The scRNA-seq data were examined employing the “Seurat” R package[19]. Quality control was executed by removing cells with greater than 5% mitochondrial gene expression. Data were corrected for batch effects between samples using the “harmony” algorithm. Data processing steps encompassed normalization, detection of genes with high variability, scaling operations, principal component analysis, neighbor identification and cluster formation. The annotation of cellular subtypes utilized established canonical marker genes, succeeded by examination of differentially expressed genes across distinct cell groups. Functional characteristics of T cells[20], macrophages, and dendritic cells (DCs)[21] were assessed using the “AddModuleScore” function. RNA-seq expression profiles from 179 ESCC cases were extracted via the GSE53625 database to examine gene expression differences between tumor and matched normal tissues.

Cell–cell interaction network analysis

The “CellChat” R package[22] was utilized to examine cell-cell communication in the TME. Analysis of ligands, receptors and related genes overexpressed in individual cell populations, integrated into a probabilistic model of communication based on known protein interaction networks, results in major afferent and efferent signals. Interactions set as having ligand or receptor expression ≥ 25% of a given cell population were considered statistically significant.

Pseudo time trajectory analysis

“Monocle“ algorithm[23] was used for sham time analysis to simulate the dynamic evolution of cell state. Cells were sorted according to gene expression similarity and cell developmental trajectories were reconstructed. Such algorithms can make predictions about the progression of cell lineages and identify critical cellular transitions in the TME.

Assessment of cell differentiation states

“CytoTRACE“ algorithm was used to assess the differentiation potential of cells. This algorithm was unsupervised and can analyze the inferred differentiation status by quantifying transcriptional diversity[24], determine the origin and development direction of individual cell subsets, especially T-cell subsets, and clarify the development level of cell subsets in tumors.

Inference of copy number variation

Copy number variations (CNVs) were analyzed using “inferCNV“ to assess genomic instability[25]. Expression data typically using T and B cells (non-malignant immune cells) are used to infer CNV of epithelial chromosomes in tumor tissue, which, based on presumed genomic changes, is used to distinguish malignant from non-malignant epithelial cells and discover features of tumorigenesis[26].

Protein-protein interaction network construction

The STRING database was employed to visualize protein-protein interaction (PPI) networks for studying the biological functions of candidate genes. Signaling pathways related to tumor progression, signaling transduction, and cell fate regulation were screened.

Gene set enrichment analysis

Differences in signaling pathway activation between different groups were assessed using gene set enrichment analysis (GSEA). The specimens were split into high and low expression groups per the median value, while Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway analyses were executed to examine the LGALS1-associated signaling cascades. GSEA was subsequently used to evaluate differential pathway activation between the two groups[27].

Patient samples and ethical approval

Ten cases of tissue specimens and clinical data were procured from patients at Sun Yat-sen University Cancer Center. All cases were confirmed by histopathological and clinical diagnoses. The study protocol was reviewed and approved by the Sun Yat-Sen University Cancer Center Ethics Committee (No. SL-B2024-312-01). Written consent was secured from participating patients, and the research methods followed the ethical guidelines established by the Declaration of Helsinki.

Cell culture

The Guangdong Esophageal Cancer Institute supplied multiple ESCC cell lines (KYSE30, KYSE150, KYSE180, KYSE410, KYSE450, KYSE510, TE-1, TE-10, and HEK293-FT). Hunan Fenghui Biotechnology Co., Ltd. furnished the normal esophageal epithelial cell line HET-1A. RPMI-1640 medium served as the culture environment for ESCC and HET-1A cells, whereas DMEM was utilized for HEK293-FT cells. The culture media comprising 100 mL/L fetal bovine serum plus 10 mL/L penicillin–streptomycin, with cell incubation occurring at 37 °C under humidified conditions with 50 mL/L CO₂.

Quantitative real-time PCR

RNA extraction was performed employing TRIzol reagent, and subsequent cDNA synthesis was conducted with the RevertAid First Strand cDNA Synthesis Kit. Real-time PCR analysis was conducted employing PowerUp SYBR Green Master Mix on a CFX96 PCR detection platform. GAPDH functioned as the housekeeping gene for normalization. The primer sequences for LGALS1 were 5′-CCTGAATCTCAAACCTGGAGAG-3′ (forward) and 5′-CACAGGTTGTTGCTGTCTTTG-3′ (reverse), and for GAPDH were 5′-CAAGAGCACAAGAGGAAGAGAG-3′ (forward) and 5′-CTACATGGCAACTGTGAGGAG-3′ (reverse).

Cell viability assay

To evaluate cell viability, cells were placed in 96-well plates at a concentration of 5 × 10³ cells/mL and subsequently exposed to cisplatin (Topscience, China) or saline solution. At 24-hour, 48-hour, and 72-hour post-treatment, cholecystokinin-8 reagent was introduced into each well, and optical density was determined at 450 nm utilizing a microplate reader.

Western blot analysis

Protein extraction was executed utilizing RIPA lysis buffer, succeeded by BCA assay quantification. The extracted proteins underwent denaturation, SDS-PAGE separation, and then transferred onto PVDF membranes. After blockade with 50 g/L skim milk, the membranes underwent overnight incubation at 4 °C with specific primary antibodies targeting key signaling and apoptosis-related proteins, encompassing β-actin, β-catenin, E-cadherin, vimentin, Bcl-2, BAX, ATR (total and phosphorylated), p53 (total and phosphorylated), Cdc25A (total and phosphorylated), and Chk1 (total and phosphorylated). The protein bands became visible through enhanced chemiluminescence after secondary HRP-conjugated antibody incubation and were documented using a Tanon 5200 imaging system.

Establishment of stable cell lines

The LGALS1-specific short hairpin RNA (shRNA) sequence was inserted into psi-LVRU6P lentiviral vector. Plasmids containing shRNA or negative control constructs were co-transfected with packaging plasmids into HEK293-FT cells to generate lentiviral particles. The viral supernatant was collected following a 72-hours incubation period and subsequently utilized to transduce KYSE30 and KYSE450 cell lines, establishing stable knockdown populations. The experimental design incorporated three distinct shRNA sequences targeting LGALS1, accompanied by a non-targeting control sequence.

Flow cytometry

Apoptosis detection was conducted utilizing the Annexin V-FITC Apoptosis Detection Kit following drug administration. For cell cycle evaluation, cells underwent fixation in 750 mL/L ethanol overnight, followed by staining with propidium iodide/RNase solution. Analyses were executed employing a CytoFLEX flow cytometer and examined through FlowJo software.

Colony formation and crystal violet staining

The 1 × 10³ cells were placed into 6-well plates, and following 14 days of cultivation, cells underwent fixation with 40 g/L formaldehyde, staining with Crystal Violet Staining Solution (Beyotime, China), and imaging. 1 × 10⁵ cells were plated into 24-well plates overnight. Complete medium containing 0 mmol/L, 2.82 × 10^-5 mmol/L, 8.47 × 10^-5 mmol/L, 2.5 × 10^-4 mmol/L, 7.6 × 10^-4 mmol/L, 2.3 × 10^-3 mmol/L, 6.86 × 10^-3 mmol/L, 2.06 × 10^-2 mmol/L, 6.17 × 10^-2 mmol/L, 1.85 × 10^-1 mmol/L cisplatin was added, respectively. Following 48 hours of treatment, cells underwent fixation with 10% formaldehyde, staining with Crystal Violet Staining Solution, and imaging.

Animal experiments

All experimental procedures underwent review and received approval from the Institutional Animal Care and Use Committee of Sun Yat-sen University (No. L102012024070D) and were executed per the Declaration of Helsinki. Utilizing three to four-week-old BALB/c-nude mice (Gempharmatech, GD, China), 20 mice were allocated into 4 groups of five animals each, with subcutaneous injection of 5 × 10⁶ tumor cells administered to each mouse. The first day of the experiment was counted after cell inoculation. Cisplatin was administered at a dose of 5 mg/kg on days 7, 14, 21 and 28. The experiment was terminated on day 32. Mice were sacrificed by neck removal, and tumor tissues were removed for subsequent experiments. Tumor sizes did not exceed 20 mm at any point, and all procedures complied with ARRIVE guidelines.

Multiplex immunohistochemistry

Tumor tissues and clinical specimens underwent formalin fixation, paraffin embedding, and sectioning. Multiplex staining was conducted utilizing the Seven-color TSA kit, with slides being scanned via the Vectra^® Polaris™ imaging system. Quantitative image analysis was conducted using Phenochart software. Antibodies targeting pan-keratin, Ki67, LGALS1, BAX, and vimentin were used to assess marker expression in situ.

Statistical analysis

Statistical evaluations and data visualizations were executed in R software (version 4.1.3). The Cox regression computations were implemented through SPSS (version 26.0). For continuous variable comparisons, the Wilcoxon test was applied, considering P values less than 0.05 as statistically significant.

RESULTS

Single-cell profiles and cell types after neoCRT in ESCC

The analysis encompassed eight ESCC specimens, comprising tumors from four neoCRT-treated patients and matched tumor-normal tissue pairs from two surgery-only patients. After quality filtering, 43407 cells met the standards, with an average of 5425 cells per specimen. Based on established markers, seven distinct cellular populations emerged: (1) T/natural killer cells; (2) Epithelial cells; (3) Myeloid cells; (4) B/plasma cells; (5) Fibroblasts; (6) Endothelial cells; and (7) Mast cells (Figure 1A and B). The samples were categorized into normal group, surgery group and neoCRT group, with analysis revealing uniform cell distribution patterns across these three groups (Figure 1C and D), demonstrating that the “harmony“ algorithm successfully minimized batch-related variations. In addition, epithelial cells, proportion and number were markedly down-regulated in the surgery group vs the neoCRT group (Figure 1E and F). Because a partial response occurred clinically in the neoCRT group, the residual epithelial cells may be related to the subpopulation that survived neoCRT treatment, which may be resistant to treatment and at greater risk for disease recurrence.

Figure 1 Comprehensive analysis of single-cell subpopulations in esophageal squamous cell carcinoma patients receiving neoadjuvant chemoradiotherapy and surgical treatment. A: UMAP visualization of 43407 high-quality cells across all 7 cellular types; B: Marker genes for distinct cell types, with coloring based on Z-score normalized expression levels; C: UMAP visualization of all cells derived from normal tissue, surgical and neoadjuvant chemoradiotherapy groups; D: UMAP visualization of all cells obtained from 8 samples; E: Distribution and quantities of distinct cell types across 8 specimens; F: Distribution and quantities of distinct cell types within the 3 groups. NeoCRT: Neoadjuvant chemoradiotherapy; NK: Natural killer.

NeoCRT treatment promoted changes in the proinflammatory microenvironment within the tumor

“CellChat“ analysis revealed a significant difference in cell-to-cell signaling between the neoCRT and surgery groups, with a substantial disparity in positive communication between fibroblasts and myeloid cells and a significant difference in negative communication between epithelial and endothelial cells (Figure 2A). Although the total number of cell-cell interactions was similar in both groups, the strength of cell-cell interactions was markedly elevated in the neoCRT group (Figure 2B). Myeloid cells showed the highest afferent signals in both conditions, indicating that myeloid cells may have a key role in ESCC (Figure 2C). NeoCRT treatment has been reported to enhance multiple inflammation-related pathways in tumors, encompassing interleukin-6, C-X-C motif chemokine ligand (CXCL), C-C motif chemokine ligand and tumor necrosis factor signaling pathways[28]. In our study, signaling pathways (epidermal growth factor, transforming growth factor-beta, platelet derived growth factor, vascular endothelial growth factor, and fibroblast growth factor) linked to tumor advancement exhibited substantial upregulation in the neoCRT group (Figure 2D), aligning with earlier observations that neoCRT treatment eliminates tumor cells while simultaneously facilitating the emergence of resistant populations[29]. Among them, the signal intensities of the proinflammatory signaling pathways CXCL, tumor necrosis factor, C-C motif chemokine ligand, and MIF were different and characterized by each other (Figure 2E).

Figure 2 Cellchat analysis demonstrates neoadjuvant chemoradiotherapy-mediated development of a pro-inflammatory microenvironment. A: Heatmap illustrating variations in interaction quantity and intensity among distinct cell populations, where red (blue) indicates enhanced (diminished) signaling in neoadjuvant chemoradiotherapy (neoCRT) relative to the surgery group; B: Comparative analysis of interaction frequencies between neoCRT and surgery groups; C: Two-dimensional representation displaying incoming interaction strengths; D: Aggregate communication probability scores for all cellular interactions in both neoCRT and surgery groups; E: Variations in inflammation-associated pathway intensity. NeoCRT: Neoadjuvant chemoradiotherapy; NK: Natural killer.

NeoCRT-induced activation of effector CD8+ T cells

Examination of 9968 T/natural killer cells revealed nine distinct subsets characterized by specific markers and functional states (Figure 3A and B). The surgical treatment group displayed higher T cell abundance (Supplementary Figure 1A), displaying notable variations in T cell subtype distribution across normal, surgical, and neoCRT specimens (Supplementary Figure 1B). Within these populations, CD8_SPRY1 cells showed prominence in normal tissue samples (Supplementary Figure 1B); existing literature links SPRY1 expression within progenitor-exhausted T cells to enhanced programmed death 1 therapeutic outcomes[30]. Surgical tumor specimens demonstrated predominance of CD8_CXCL13 and Treg populations expressing PDCD1 and CXCL13 (Supplementary Figure 1B). The presence of Treg infiltrates indicates unfavorable prognosis[31], while elevated CXCL13-positive T cells, marked by exhaustion indicators and diminished effector molecules, facilitate immune evasion[32,33]. Tumors without treatment consequently harbored a greater abundance of immunosuppressive cells. Conversely, CD8+ T cells that express cytotoxic molecules (GZMK, GZMA, GZMB, and NKG7) demonstrated significant upregulation in the neoCRT group, suggesting that neoCRT treatment enhanced antitumor immune activity. The surgery group exhibited higher exhaustion scores, while the neoCRT group demonstrated elevated cytotoxicity and interferon response scores (Figure 3C). Pseudo time trajectory examination revealed CD8+ T cell transformation from proliferative to effector states, ultimately resulting in CD8+ T cell depletion (Figure 3D and E). Subgroup examination demonstrated that activated CD8+ T cells predominantly accumulated during initial differentiation following neoCRT treatment (Figure 3F and Supplementary Figure 1C and D). The “CytoTRACE“ assessment validated these findings, ranking the CD8_proliferation subset with maximal differentiation potential, succeeded by effector and depleted conditions. The CD8_SPRY1 subset exhibited minimal differentiation capacity, potentially reflecting its presence in normal tissues where effector T cell requirements remain minimal (Figure 3G-I). The capacity of neoCRT to effectively enhance anti-tumor immune responses within the tumor environment demonstrates its therapeutic significance in locally advanced ESCC.

Figure 3 Neoadjuvant chemoradiotherapy facilitated CD8+ T cell activation. A: UMAP visualization displaying 9 distinct clusters comprising 9968 T/natural killer cells; B: Heatmap illustrating marker gene expression across T cell subsets; C: Comparative analysis of T-cell functional scores; D: Temporal trajectory analysis of the entire CD8+ T cell population; E: Positioning of CD8+ T cell subtypes along the cellular developmental trajectory; F: Distinct groups along the cellular developmental trajectory; G: CytoTRACE score distribution throughout cellular developmental pathways; H: UMAP representation of CytoTRACE score distribution; I: Comparative differentiation levels among various CD8+ T cell subsets. ^dP < 0.0001. IFN: Interferon; Neo: Neoadjuvant.

NeoCRT boosted the anti-tumor activity of macrophages

Among 3013 myeloid cells, the cluster was separated into five monocyte and macrophage subsets, three DC subsets, and a proliferative cell cluster, with myeloid cells being more abundant in the neoCRT group (Figure 4A-C). Macro_TREM2, a subset of myeloid cells that accounted for ≥ 75% in the surgery group, has been reported to inhibit antigen presentation, reduce T-cell infiltration, and be associated with poor prognosis in a variety of cancers[34]. NeoCRT treatment markedly diminished Macro_TREM2 infiltration, replacing it with a diverse monocyte/macrophage subpopulation (Figure 4D). Using classical M1/M2 molecular markers, we identified Mono_S100A9 as an early monocyte population lacking typical macrophage features. Macro_ISG15 displayed an M1-like inflammatory profile, whereas Macro_TREM2 aligned with an M2-like phenotype. Interestingly, a subpopulation co-expressing S100A2 and TREM2 produced inflammatory mediators such as CXCL3 (Figure 4E), suggesting a hybrid M1/M2 state consistent with recent studies reporting discordance between TREM2 expression and classical macrophage polarization[35]. The DC categories included DC_LAMP3, plasmacytoid DCs, and myeloid DCs. Myeloid DCs showed abundance in normal tissues but experienced differentiation throughout tumor development (Figure 4F). Functional characterization demonstrated that Macro_TREM2 primarily engaged in proliferation and tissue restoration, whereas Macro_S100A2 was more involved in transforming growth factor-beta, vascular endothelial growth factor signaling, and extracellular matrix restructuring (Figure 4G and H). Functionally, DC_LAMP3 constituted the most activated DC subset, regulating tumor-infiltrating T cells through co-stimulatory molecules and migrating to lymph nodes[36]. Plasmacytoid DCs, which produce interferon when responding to foreign elements, were frequently inhibited by immunosuppressive components in the TME, linking to unfavorable prognosis[37].

Figure 4 Characteristic alterations in myeloid cells following neoadjuvant chemoradiotherapy. A: UMAP visualization of 9 distinct clusters comprising monocytes, macrophages, and dendritic cells; B: Marker genes identifying the 9 clusters; C: UMAP visualization depicting the distribution pattern of myeloid cells across diverse groups; D: Compositional analysis of 9 myeloid cell subsets across diverse groups; E: M1 and M2 signature scoring of monocytes and macrophages; F: Activation, migration, and tolerance signature scoring of dendritic cells; G: Kyoto Encyclopedia of Genes and Genomes pathway enrichment in distinct monocyte and macrophage populations; H: HALLMARK pathway enrichment in distinct monocyte and macrophage populations. NeoCRT: Neoadjuvant chemoradiotherapy.

Remaining tumor cells following CRT exhibit potential resistance characteristics

After analyzing 17356 epithelial cells, it was observed that the neoCRT group contributed predominantly (95.3%) to the sample population (Supplementary Figure 2A and B). Since patients receiving NeoCRT treatment exhibited PR clinically, the persisting tumor cells potentially demonstrated resistance characteristics. The CNV analysis revealed that NeoCRT treatment resulted in minimal survival of normal epithelial cells (Figure 5A-C and Supplementary Figure 2C and D). After the elimination of normal epithelial cells from consideration, comparative analysis between malignant epithelial cells from the neoCRT group and surgery group identified differential genes associated with inflammatory response, cell adhesion, antigen presentation, and immunity (Figure 5D-F). The PPI network of these differentially expressed genes was constructed and 20 key genes with high connectivity were shown. After excluding HLA and cytokeratin family members, we identified resistance related genes, such as EGFR, LGALS1, LCN2 and FABP5 (Figure 6A). Through examination of the GSE53625 dataset, the analysis revealed elevated LCN2 and FABP5 expression in normal tissues, leading to their exclusion (Figure 6B). Considering numerous EGFR-targeted preclinical investigations[38], LGALS1 emerged as the primary target regarding neoCRT resistance in ESCC. LGALS1, belonging to the galectin protein family, demonstrates connections with tumor development, immune system avoidance, resistance to apoptotic processes, and blood vessel formation[39], correlating with unfavorable outcomes in malignancies[40]. Additional pseudotime evaluation of malignant epithelial cells suggested that post-neoCRT residual cells predominantly appeared during initial developmental phases (Figure 6C and D). The LGALS1 expression pattern manifested in early stages, suggesting potential involvement in neoCRT resistance mechanisms (Figure 6E and F). Given that patients underwent neoCRT protocols incorporating radiation treatment and platinum-based chemical agents, subsequent investigations examined LGALS1’s role in platinum resistance.

Figure 5 Detection of malignant epithelial cells. A and B: The inferCNV algorithm was employed to infer malignant epithelial cells in the neoadjuvant chemoradiotherapy and surgery groups; C: UMAP visualization depicting the spatial distribution of malignant and normal epithelial cells; D: Differentially expressed genes in malignant epithelial cells between neoadjuvant chemoradiotherapy and surgery groups; E and F: Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analyses of the differentially expressed genes. NeoCRT: Neoadjuvant chemoradiotherapy; NK: Natural killer.

Figure 6 Detection of essential drug resistance-associated genes. A: Protein-protein interaction networks for differentially expressed genes; B: Expression of key genes in tumor and normal tissues from GSE53625; C: LGALS1 expression across all epithelial cells; D: Pseudotime analysis of all malignant epithelial cells, with red indicating the starting point of cell development; E: Distribution of neoadjuvant chemoradiotherapy and surgery groups along cell developmental trajectories; F: Distribution of LGALS1. ^dP < 0.0001. NeoCRT: Neoadjuvant chemoradiotherapy.

LGALS1 augmented cisplatin resistance in ESCC in vitro

Bioinformatics analysis revealed LGALS1 exhibits elevated expression patterns in tumor tissues, suggesting its role as an oncogenic factor. Subsequent experimental validation examined LGALS1 functionality through in vitro studies. The application of multiplex immunohistochemistry techniques enabled detection of LGALS1, Ki67, and Pan-keratin expression levels across ESCC and normal esophageal tissue specimens. The simultaneous presence of Ki67 and Pan-keratin markers in ESCC samples facilitated tumor tissue identification. Analysis indicated that LGALS1 demonstrated marginally reduced expression in ESCC compared to normal esophageal tissue samples (Figure 7A). Western blot and quantitative real-time PCR confirmed variable LGALS1 expression across ESCC cell lines and normal epithelial cells (Supplementary Figure 3A and B). Treatment of KYSE30 and KYSE450 cells with cisplatin induced upregulation of LGALS1 (Figure 7B and C). Stable knockdown cell lines using shRNA targeting LGALS1 were established (Supplementary Figure 3C and D), selecting shRNA1 for KYSE30 and shRNA2 for KYSE450 for downstream assays. Neither knockdown nor control cells showed significant differences in proliferation or morphology under basal conditions (Figure 7D and E). However, upon cisplatin treatment, LGALS1 knockdown cells exhibited increased sensitivity, as demonstrated by cholecystokinin-8 viability assays and flow cytometry, indicating that LGALS1 confers resistance by mitigating cisplatin cytotoxicity (Figure 7F and G). PPI network and enrichment analyses linked LGALS1 to EMT, cell adhesion, growth signaling, and cytokine pathways, aligning with established findings (Supplementary Figure 4A and B)[39]. GSEA of bulk RNA-seq from 179 ESCC patients additionally correlated elevated LGALS1 expression with apoptosis and DNA repair pathways, reinforcing its involvement in cisplatin resistance (Supplementary Figure 4C). Western blot demonstrated that LGALS1 knockdown paired with cisplatin treatment diminished phosphorylation of DNA repair proteins ATR, CDC25A, Chk1, and p53 (Figure 7H). EMT and apoptosis markers likewise altered, with cisplatin treatment decreasing vimentin and Bcl2 while elevating β-catenin, E-cadherin, and BAX expression (Figure 7I). Crystal violet staining verified that LGALS1 knockdown inhibited cell proliferation and amplified cisplatin cytotoxicity (Figure 7J). Collectively, these findings demonstrate that LGALS1 downregulation potentiates cisplatin efficacy by inhibiting EMT and DNA repair mechanisms.

Figure 7 LGALS1 diminishes esophageal squamous cell carcinoma cell responsiveness to cisplatin treatment. A: LGALS1, Ki67, and Pan-keratin expression levels in esophageal squamous cell carcinoma specimens compared with adjacent normal esophageal tissue samples; B: Assessment of LGALS1 expression changes under cisplatin administration utilizing quantitative real-time pathological complete remission methodology; C: Evaluation of LGALS1 protein levels following cisplatin exposure as determined by western blot analysis; D: The influence of LGALS1 expression on the cell cycle was analyzed by flow cytometry; E: Clone formation assay examined the influence of LGALS1 expression on cell proliferation; F: Cholecystokinin-8 examined the influence of LGALS1 expression on tumor cell viability under cisplatin treatment; G: The influence of LGALS1 expression on tumor cell apoptosis was ascertained by flow cytometry with cisplatin treatment; H: The phosphorylation levels of the DNA repair-related proteins ATR, CDC25A, Chk1, and p53 were measured by western blot; I: The expression of epithelial-mesenchymal transition and apoptosis-related proteins beta-catenin, E-cadherin, Vimentin, and Bcl2 and BAX was measured by western blot; J: Crystal violet staining was utilized to detect LGALS1 expression on tumor cell proliferation by cisplatin treatment with a concentration gradient. ^aP < 0.05, ^bP < 0.01, NS: Not significant. ESCC: Esophageal squamous cell carcinoma.

LGALS1 suppresses cisplatin-induced elimination of ESCC tumors in vivo

To assess LGALS1’s function in cisplatin resistance within an in vivo context, LGALS1 knockdown and control KYSE30 cells were subcutaneously transplanted into BALB/c nude mice. Mice received intraperitoneal injections of cisplatin (5 mg/kg) following a predefined schedule, after which tumors were harvested (Figure 8A and B). Tumor volumes showed no significant differences between groups when treatment was absent. However, cisplatin-treated mice bearing LGALS1 knockdown tumors exhibited markedly diminished tumor volumes compared to controls (Figure 8C-E). Multiplex immunohistochemistry analysis of tumor tissues revealed that although LGALS1 was knocked down in tumor cells, LGALS1 expression was elevated in the tumor stroma after cisplatin treatment, presumably owing to induction in basal cells and cancer-associated fibroblasts. Notably, LGALS1 was more abundant in tumor cells of control mice. Cisplatin treatment elevated BAX expression while reducing Ki67 and vimentin levels in LGALS1 knockdown tumors, indicating enhanced apoptosis and reduced proliferation and EMT. Differences between control and treated groups were less marked in tumors with intact LGALS1 expression, likely reflecting drug resistance. These data suggest that LGALS1 depletion sensitizes ESCC tumors to cisplatin by diminishing resistance mechanisms in vivo.

Figure 8 LGALS1 diminishes the cisplatin sensitivity of esophageal squamous cell carcinoma CDX model. A: Experimental protocol for subcutaneous tumor formation and dosing experiments in mice; B: Mice and individual tumors after different treatment approaches; C: Statistical plots of tumor weights in mice following diverse treatments; D: Growth curves of mice following various treatments; E: The LGALS1, Ki67, and BAX and vimentin protein were detected in tumor sections by multiplex immunohistochemistry staining. ^bP < 0.01, NS: Not significant.

DISCUSSION

Multimodal therapy has markedly enhanced outcomes for patients with locally advanced ESCC, with neoCRT followed by surgery currently established as the standard of care. While neoadjuvant immunotherapy is emerging, its effects on the TME remain under investigation. Previous scRNA-seq investigations have predominantly focused on immune cell dynamics after neoadjuvant treatment but have often overlooked tumor-intrinsic resistance mechanisms[29,30,41].

In this investigation, scRNA-seq was employed to comprehensively characterize the TME and residual tumor cells in ESCC patients post-neoCRT. Our analysis revealed a profoundly activated TME marked by upregulation of inflammatory signaling pathways and increased infiltration of effector CD8+ T cells, which underscores neoCRT’s capacity to invigorate anti-tumor immunity. At the same time, we found significant remodeling of myeloid compartments, a decrease in immunosuppressive TREM2-positive macrophages, and an increase in inflammatory macrophage subsets and LAMP3+ DCs. Among these, LAMP3+ DCs serve a vital function in antigen presentation and T cell activation.

Although neoCRT treatment can produce many beneficial immune responses to promote tumor extinction, there are still malignant epithelial cells surviving after treatment, which poses a major potential risk for future metastasis and recurrence. LGALS1 was found to be highly expressed in residual malignant epithelial cells and associated with neoCRT resistance. The known roles of LGALS1 in tumor proliferation, immune evasion, apoptotic resistance, and angiogenesis are consistent with the results of our bioinformatics analysis[42-46]. Functional analysis demonstrated that LGALS1 knockdown enhanced the sensitivity of ESCC to cisplatin treatment in vitro and in vivo through suppression of EMT, DNA repair, and anti-apoptotic pathways. The results of functional identification experiments were consistent with those reported in other tumors. LGALS1 mediated chemoresistance through EMT and autophagy pathways[47-49]. In addition, LGALS1 seems to contribute to the establishment of immunosuppressive microenvironment in tumors. Elimination of tumor-associated macrophages has the ability to reduce LGALS1 expression and enhance CD8+ T cell infiltration, suggesting that LGALS1 serves a dual function in tumor cell intrinsic drug resistance and immune-mediated drug resistance[50].

However, the research still has limitations. The relatively limited sample size and absence of data on patients who achieved pCR restrict the generalizability of these findings. Of course, it was very difficult to collect data for pCR. The paucity of viable tumor cells in the pCR data also limits validation. Future studies should further dissect the molecular mechanism of LGALS1-mediated cisplatin resistance and explore therapeutic strategies targeting LGALS1. Given the current clinical challenges of direct LGALS1 inhibition, investigating combination therapies combining LGALS1 inhibition with standard therapies may provide a promising approach to overcome drug resistance in ESCC.

CONCLUSION

This investigation reveals crucial understanding regarding neoCRT’s effects on the TME characteristics among locally advanced ESCC patients through single-cell analysis. Importantly, we identified LGALS1 as a key molecule associated with drug resistance, highly expressed in residual tumor cells following treatment. Targeting LGALS1 presents a promising strategy to overcome platinum resistance and enhance therapeutic efficacy in ESCC.

ACKNOWLEDGEMENTS

The authors sincerely appreciate the staff of the Guangdong Esophageal Cancer Institute for completing the project.